brand-generator

Brand Namer

Generate unique brand names with verified-available domains through structured discovery and linguistic construction.

Workflow

Phase 1: Discovery Interview

Conduct a focused interview to understand the product before generating any names. Ask these questions across 1-2 messages (do not overwhelm with all at once):

Round 1 — Core understanding:

• What does the product/company do in one sentence?
• Who is the target user?
• What category is this — a tool, platform, AI agent, marketplace, or consulting brand?

Round 2 — Strategic constraints:

• Will this expand beyond its current scope? (Determines how abstract the name should be)
• What trust level is needed? (Career/finance/legal = serious tone; consumer/social = more flexibility)
• What tone do you want? (e.g., technical, warm, minimal, bold, playful)
• Any words or themes to avoid?
• Preferred domain TLD? (.com ideal, .ai and .co acceptable)

Do NOT proceed to name generation until discovery is complete.

Phase 2: Name Generation

Read references/naming-framework.md for construction patterns, phonetic rules, risk filters, and the quality scoring matrix.

Generate 20-25 candidate names by:

1. Brainstorm semantic roots on the fly — Do NOT rely on a static word list. Derive novel roots from the product's specific domain using: metaphors, foreign language fragments (Latin, Greek, Japanese, etc.), cross-domain analogies, phoneme invention, and abbreviation blends. Aim for roots that are not common English dictionary words.
2. Apply construction patterns (root + suffix, truncation, blending, vowel shifts, letter removal, partial root blends)
3. Apply phonetic rules — 2-4 syllables, one obvious pronunciation, passes phone test
4. Pre-filter for novelty — Before any domain check, reject candidates that are recognizable English words, obvious brand roots, or that you estimate are highly likely to be taken (e.g., "fitly", "formio", "pathly")

Phase 3: Domain Verification

Check domains for all surviving candidates in a single batch using the bundled script:

scripts/check_domain.sh name1.com name1.ai name1.co name2.com name2.ai name2.co ...

Pass all candidates × all TLDs (.com, .ai, .co) in one call. Reject candidates where all reasonable TLDs are taken.

Phase 4: Score & Web Validate Survivors

Only now score and validate — don't waste effort on names with no available domains.

1. Score each surviving candidate on the 6-dimension quality matrix (reject below 20/30)
2. Web validate the top scorers via WebSearch (batch 3-4 searches in parallel):
   • No dominant company on SERP page 1 for the name
   • Not a medical/pharma term
   • Not a foreign-language profanity
   • No obvious trademark conflicts in the same industry
   • Reject any name that surfaces conflicts

Phase 5: Present Shortlist

Present 3-5 finalists in this format:

## [Name] — [domain.tld]

- **What it evokes:** [brief description of the associations]
- **Score:** [X/30] (Distinctiveness / Pronounceability / Memorability / Flexibility / Legal Safety / Brandability)
- **Domain status:** Available ([tld])
- **Risk notes:** [any minor concerns or "Clean"]

After presenting, ask the user which direction resonates and offer to:

• Refine around a favorite
• Generate more in a specific style
• Run deeper validation (USPTO TESS, Crunchbase, GitHub, LinkedIn, social handles)

Phase 6: Deep Validation (on request)

When the user narrows to 1-2 finalists, run deeper checks via WebSearch:

• "[name]" site:crunchbase.com
• "[name]" site:github.com
• "[name]" site:linkedin.com/company
• "[name]" trademark software
• "@[name]" twitter / "@[name]" instagram

Report findings and confirm the final selection is clear.